Explainer

WiFi sensing as a people counter

Ordinary WiFi signals can be used to detect, count, and estimate the density of people in a space — without cameras, wearables, or an app on anyone’s phone. Here’s how it works, and why it matters.

The physics, briefly

Every WiFi router fills a room with radio waves. Human bodies are mostly water, so they absorb, reflect, and scatter those waves in measurable ways. Modern WiFi chipsets expose Channel State Information (CSI) — a per-subcarrier record of how the signal was distorted between transmitter and receiver. Movement, breathing, and even a still person sitting in a chair each leave a distinct fingerprint in CSI.

Feed a stream of CSI samples into a small neural network and you can recover: is anyone here, how many, are they moving, and — with enough antennas — rough pose and location. Open-source projects such as RuView run this pipeline on $9 ESP32 boards.

As a people-density counter

For crowd analytics, the interesting output isn’t identity — it’s occupancy. A CSI-based counter can report:

Number of people in a zone, updated in real time.
Entries and exits across a doorway.
Dwell time and queue length.
Per-room density in a building, without per-room cameras.

Typical accuracy on published benchmarks is around 80–95% for counts up to 5–10 people per sensor, depending on room geometry and calibration time (usually 30 seconds of empty-room baseline).

Where it’s already used

Smart-building HVAC that scales with actual occupancy.
Retail foot-traffic and queue analytics.
Aged-care monitoring: fall detection, room-inactivity alerts.
Emergency-response headcounts during evacuations.
Research: contactless breathing and heart-rate estimation.

Privacy implications

WiFi sensing is often marketed as “privacy-preserving because there’s no camera.” That’s only partly true.

It works through walls and in the dark. Someone outside a room can infer what’s happening inside without any visible sensor.
No opt-in signal. Unlike a camera or a Bluetooth beacon, a passive CSI receiver leaves no trace users can detect.
Re-identification is possible. Gait, body shape, and even keystroke patterns have been recovered from CSI in academic work. “Anonymous count” is a design choice, not a physical limit.
Regulatory grey zone. Australian privacy law (Privacy Act 1988, APPs) covers “personal information.” Aggregate counts arguably fall outside; individual tracking almost certainly does not.

Why OzWatch doesn’t map it

Unlike ANPR cameras or cell towers, WiFi CSI sensors have no public registry. They’re usually indistinguishable from a regular WiFi access point, and any modern router with the right firmware can become one. Mapping them accurately isn’t possible today; we link to this explainer instead so the capability is at least visible.